On Tuesday 19 October 2010, Steven Rostedt wrote:
> On Tue, 2010-10-19 at 15:36 +0200, Arnd Bergmann wrote:
> > [trimming Cc list]
> >
> > On Tuesday 19 October 2010, Steven Rostedt wrote:
> > > I think we also need to cover the PREEMPT case too. But that could be a
> > > compile time check, since you can't boot a preempt kernel and make it
> > > non preempt.
> >
> > Right. Can we turn the lock_kernel() into preempt_disable() in these
> > drivers when we know we never run on SMP?
>
> I'm not sure that will work. A holder of the BKL can call schedule or
> even a mutex. The schedule code will drop the BKL and re-enable
> preemption. Unless the code is known not to schedule while holding BKL,
> we would need to open code the preempt_enable() around the locations
> that the code may schedule.
Right, that won't work then. I was confused by the fact that __lock_kernel()
turns into preempt_disable() on UP-preempt systems, which only works
because it still maintains current->lock_depth and schedule consequently
enables preemption again in __release_kernel_lock(). With CONFIG_BKL
disabled, that won't happen any more.
Arnd
